About
I'm a serial founder building at the intersection of AI and societal impact. I'm driven…
Articles by Avni
Activity
4K followers
-
Avni Patel Thompson shared thisI was honoured to speak with the Mayo Clinic Nursing team leading the charge on what transformation looks like, informed by the people "walking the land" of everyday challenges. Each of us is faced with a tremendous amount of change and choice today and I had the chance to share how I believe the humble cow path gives us a lens by which we can intentionally and ambitiously preserve the good friction that creates texture while aggressively addressing the limiting friction that is burden.Avni Patel Thompson shared thisLast week, Avni Patel Thompson spoke to our #NursingNexus group, and one idea immediately took me back to a very different season of my life: the invisible load we carry. For me it was three kids at three different schools. What felt like 50 newsletters and emails a week, filled with information I was somehow supposed to sort, remember, and connect to the right child. I'll admit it—I eventually gave up. My kids were in middle and high school, and I told them they were going to have to keep track of it themselves. Avni experienced that same kind of friction as a parent and built a solution for it. I would have gladly handed over some of that mental load—and the guilt that came with never feeling quite on top of it. Her story made me think about the work we're already doing in nursing at Mayo Clinic. How much invisible load have we accepted because "that's just how it works"? We're actively working to remove that burden and redesign how work gets done, but we need to keep going. Faster. More boldly. And differently than we have before. Avni shared several ideas that gave me new language for that work: think like a scientist. Learn to use friction. Solve the problems you live. Iterate, even when it's hard. Her point that friction has texture and contours, rather than simply being a barrier to remove, especially resonated with me. What I took away from her talk was bigger than technology. It was a challenge to be intentional about the future we're building for ourselves and others. At Mayo Clinic Nursing, we're already building that future. Now we need to accelerate—keep questioning what we've accepted, learn from the friction, and create something better for the patients we serve. #MayoClinic #NursingInnovation #HealthcareInnovation
-
Avni Patel Thompson shared thisOur AI is missing a critical kind of intelligence - Relational Intelligence. Reasoning that is anchored in the beauty and complexity of our relationships. Decisions that show up on the surface but are determined by invisible, interwoven threads deep below. The trouble is, without teaching our AI models the shape of this reasoning, it leaps to the patterns it does know - that of logical, transactional, single answer reasoning. Anne-Marie Slaughter and I dig into why that's a problem and what we can do about it. Because so many of the big challenges of our time rely on our relational threads to be acknowledged and reinforced as we build. But if we can do this right - if we can build in Relational Intelligence so that we can lean into the intentional texture - we can finally start solving these previously intractable challenges. https://lnkd.in/gfsmUe_J
-
Avni Patel Thompson shared thisYou can't model what you haven't seen. LLMs are pattern recognition machines trained largely on the text of the public internet, reinforced for single-turn (re: immediate) helpfulness. That's incredibly powerful for verifiable domains where text can fully convey the value and the data sets contain wide range of cases. But this isn't the case for physical intelligence. Text can't capture the know-how of how to grip a mug, no matter how detailed the description. And so we see the investment and the progress in organizations inventing the sensors, collecting the data and training the models. This is the same of Relational Intelligence - much of our lives is governed by reasoning tied to our relationships. But they often masquerade as logistical questions and so our current AI takes what it knows and confidently starts answering. To understand why this is problematic and how it will limit our success in applying AI in domains like education, healthcare and even customer service, we can look at this simple example of asking ChatGPT and Claude about how to navigate a single calendar conflict. The response is totally plausible and reasonable. The problem is: a) both Chat GPT and Claude assumed my family's values b) both assumed the value was "honor commitments" as if it were the only one c) both attempted to optimize both by suggesting a "swing by" to the birthday party Instead, AI built with relational intelligence needs to be: 1) longitudinal - solve over time 2) plural - allow for optimizing multiple (sometimes conflicting) values 3) boundaried - understand where the AI nudges the user to lean into the relationship But first? To do any of this we need benchmarks. We need to measure how our current SOTA models do along relational reasoning and create a bar of "what great looks like" so we can start the hill-climbing. In the meantime, we need to notice - when are we asking questions of our AI helpers that are relational in nature? What are the responses and how do they differ from how we humans might actually reason through them? I'm collecting examples so I'd love to see yours.
-
Avni Patel Thompson shared thisThanks TED Conferences for inviting me to share some thoughts on how we each need to build our own intuition around our desired role of tech in our lives. But to do that - to know where that line is -we each need to engage with the line. It’s only by crossing the line that we really know it’s there so I think we need to find ways of playing and experimenting that feels responsible and fun and easy.Avni Patel Thompson shared thisNew technologies — like AI, drones, robotics and more — are accelerating at a rapid pace and it will take all of us to get it right. Today at TED2026, Renae explores the big questions we are asking around this tech and how we as humans can grow alongside them as we head into this exciting, yet uncertain, new future:Overwhelmed by AI? You have more power than you thinkOverwhelmed by AI? You have more power than you think
-
Avni Patel Thompson reposted thisAvni Patel Thompson reposted thisNew technologies — like AI, drones, robotics and more — are accelerating at a rapid pace and it will take all of us to get it right. Today at TED2026, Renae explores the big questions we are asking around this tech and how we as humans can grow alongside them as we head into this exciting, yet uncertain, new future:Overwhelmed by AI? You have more power than you thinkOverwhelmed by AI? You have more power than you think
-
Avni Patel Thompson posted thisOpenclaw 🦞, Claude code 👾, Hermes ... since January we've seen huge breakthroughs in how agents are architected and deployed, over the simplest channels like Whatsapp and Telegram. But it's a good news/bad news kind of thing: - good because these approaches are giving the power to the people to build the very useful things themselves - bad because man, it's still so power user "read the docs" oriented After wading in for a bit with Claude Code I decided if I was really going to understand the advances I had to dive in completely. So I got myself a Mac mini and spun my my own OpenClaw 🦞 agent. And wow - there's a lot to be worked out, but I can tell you, after working on building a family version of this for 4 years, this is total breakthrough. It enables a level of individual utility with a set of simple markdown files as the "harness". But it hasn't been until I started combining some of my life/work best practices that it's really gone from "toy" to "daily helpful". Just one example is sharing my quarterly planning approach and working with it to refine my goals and helping to break things down into monthly/ weekly goals and holding me accountable. The experience has showed me just how much I think we need a trusted way to traverse the frontier of AI right now because this is where all the good stuff is. I really want others to experience what this feels like but it's taken a lot of trial and error, reading a ton of docs and constant optimizing on cost and security. Curious if others are experimenting with personal agents on Claude or Openclaw and what's been an unlock? Or if you're curious and what is the biggest barrier?
-
Avni Patel Thompson shared thisThis has been swirling inside for me, building for months and now, years. This feeling and belief that technology isn't the problem, but the way that we're hellbent on building it in only one flavor, is. For big and more and convenient and efficient. Outsized gains going to the few because of a system they think is clever but really is just lazy and extractive. But what about all of the spaces and people who might want... less? meaningful? Quiet. The past couple of months have been so exciting because I see there are many of us asking these same questions. But because we're all building for quiet, it's been hard to know that there are others with the same beliefs. I think it's time to build the right kind of beacon for a different kind of tech. Quiet tech. cc: Anne-Marie Slaughter Alex Komoroske Aishwarya Khanduja Elizabeth Dorman Laura Danger Anjan Katta Jean Hsu Reni Cao Hallie Horvath https://lnkd.in/gaZA3gpk
-
Avni Patel Thompson reposted thisAvni Patel Thompson reposted thisWe just released some insights from testing 32 AI models for child safety. The 3 most interesting to me are: 1. Models are not getting safer over time (see chart below). 2. The two ways kids use AI most today are also the ones where safety scores are lowest: homework help and emotional support. 3. Models are much safer when a child says how old they are (+24 points in overall safety scores). Thank you Eleni Pavlakis, Marc Sarfati, Stéphie Herlin and Cyril Gantzer for your help with this ❤️ https://lnkd.in/grmuhreu
-
Avni Patel Thompson shared thisThis is one of the best articulations I've seen on the divide. But I'm also not surprised. These LLMs were quite literally designed and built for purposes on advancing science and computation, something coding aligns well with. The dominant benchmarks have been designed and used to improve the models along these specific capabilities. So imagine what might be possible if we had purpose built models that supported properly taking away the drudgery of so many other jobs. Jobs where it's the bureaucracy and drudgery that fogs up and dims the human magic. Basically what this respondent to Anthropic's recent sentiment survey said: https://lnkd.in/gb5cJmk5 “AI should be cleaning windows and emptying the dishwasher so I can paint and write poetry. Right now it’s exactly the other way around.” We need to first imagine what that might look like and then get to work building it. Margin challenged and resource constrained spaces like teaching and nursing and childcare but also bookstores and personal training and local coffee shops - imagine of the people were freed up to do their superpowers while the computers handled the things that made it possible to be a one-person powerhouse. It doesn't feel right because it isn't right. Yet. These machines were never built for these spaces but that doesn't mean they can't be.Avni Patel Thompson shared thisThere are lots of reasons to be concerned about AI displacing tech jobs. But this quote speaks, quite elegantly, to why many in the industry are optimistic about what it could mean for them. From an excellent Clive Thompson piece: https://lnkd.in/eiRWuQuZ
Avni Patel Thompson replied to a comment
5mo
Kurt Wellington I totally agree but if you can’t access the set up to be able to share your goals and insights easily and effectively, you still can’t access the power.
Luckily I think this is where we’ll see lots of progress in the next 6 months
Avni Patel Thompson replied to a comment
6mo
Haley Swenson Thank for flagging - would love to explore this topic and our findings with this group. I agree that we're not designing/optimizing AI for our collective futures: 2/3 of americans live in a family construct; while all of our live in neighborhoods. What does AI designed to consider the collective goals look like? What are gaps from getting us there (eg. benchmarks, the right training data, and the right reinforcement processes).
-
Avni Patel Thompson reacted on thisAvni Patel Thompson reacted on thisDid I ever think when I became a nurse that I would need to understand data architecture and vectors? Absolutely not. Yet here I am. I've always valued learning, but I'm realizing that a growth mindset asks something more of us: being willing to become a beginner again, even after we've spent years becoming an expert. That's not always comfortable. The more senior we become, the easier it can be to stay in the spaces where we know the language and have the answers. But healthcare is changing too quickly for any of us to do that. I don't need to become a data architect. But if I can understand even 30% of the language, concepts, and implications, I can ask better questions, make better decisions, and lead differently because of what I'm learning. Expertise may have helped us become leaders. The willingness to become beginners again may be what helps us lead what's next. So yes, apparently this nurse is learning about vectors. 😊
-
Avni Patel Thompson reacted on thisAvni Patel Thompson reacted on thisLast night I was inducted into the Dartmouth Entrepreneurs Hall of Fame. Dartmouth College changed the trajectory of my life. I was a kid from Boise who was not the obvious candidate, someone in admissions took a chance on me, and everything since traces back to that decision. I am so grateful. Thank you to Trevor Jensen and Maia Josebachvili for the introduction and for friendships that now span decades. To Jamie Coughlin for his leadership, to President Sian Leah Beilock, and to the Magnuson Center for Entrepreneurship at Dartmouth for making evenings like this possible. And to my parents, Skip Oppenheimer and Esther Oppenheimer, who were in the room last night the way they've been in my corner my whole life. It's humbling to join a group that includes Shonda Rhimes, Jim Coulter, Steve Mandel, Mark Mader, and Brent Frei. Special thanks to Mark Mader for being there last night. My full acceptance remarks are in the article below... about the strengths that built Remitly, and the shadows they cast.
-
Avni Patel Thompson reacted on thisAvni Patel Thompson reacted on thisLast week, Avni Patel Thompson spoke to our #NursingNexus group, and one idea immediately took me back to a very different season of my life: the invisible load we carry. For me it was three kids at three different schools. What felt like 50 newsletters and emails a week, filled with information I was somehow supposed to sort, remember, and connect to the right child. I'll admit it—I eventually gave up. My kids were in middle and high school, and I told them they were going to have to keep track of it themselves. Avni experienced that same kind of friction as a parent and built a solution for it. I would have gladly handed over some of that mental load—and the guilt that came with never feeling quite on top of it. Her story made me think about the work we're already doing in nursing at Mayo Clinic. How much invisible load have we accepted because "that's just how it works"? We're actively working to remove that burden and redesign how work gets done, but we need to keep going. Faster. More boldly. And differently than we have before. Avni shared several ideas that gave me new language for that work: think like a scientist. Learn to use friction. Solve the problems you live. Iterate, even when it's hard. Her point that friction has texture and contours, rather than simply being a barrier to remove, especially resonated with me. What I took away from her talk was bigger than technology. It was a challenge to be intentional about the future we're building for ourselves and others. At Mayo Clinic Nursing, we're already building that future. Now we need to accelerate—keep questioning what we've accepted, learn from the friction, and create something better for the patients we serve. #MayoClinic #NursingInnovation #HealthcareInnovation
-
Avni Patel Thompson reacted on thisThis is everything. I had tears in my eyes and chills all over my body when I read the handwritten note. I am so grateful that our our Wildland fire helmet saved Landen's life and allowed him to marry the love of his life, Hollie! Here's to a lifetime of health and happiness! Thank you for trusting Bullard! And thank you for taking the time and care to share your beautiful story with us!Avni Patel Thompson reacted on thisA wedding invitation arrived at Bullard HQ with a handwritten note. Landen Bender is a wildland firefighter. Just over a year ago, he was crushed by a tree while on the job. In his own words: "If it weren't for one of your helmets, I wouldn't have a wedding at all. If it wasn't for your helmet, I'd be dead." Landen married his wife Hollie on July 18, 2026. He sent us the invitation to say thank you. This is why we do what we do. The shifts that go on. The lives that continue. The weddings that happen because someone made it home. We're honored to welcome Landen into the Bullard Save-A-Life Club, and we wish him and Hollie a lifetime together.
-
Avni Patel Thompson reacted on thisAvni Patel Thompson reacted on thisLast Friday I wrapped up my second tour at BCG and this is week one for me in a newly-created VP of Data Solutions role at John Deere. Our charge: ship data and AI capabilities to the farmers and builders who run on Deere iron around the world - helping them improve profitability, productivity, and sustainability. This is my third run at the crossroads of agriculture and data, but this time it sure feels different - the sheer scale of data that's in play, what GenAI is doing to our ability to translate all of that into solutions, and a market that seems primed for something new. Hard to imagine a better moment to tackle this and we're growing an exceptional cross-functional team - if the mission and the moment feel like a fit, be in touch! I'm grateful to my colleagues at BCG for two unforgettable runs, and I'm delighted to be rolling up sleeves alongside Jahmy Hindman and his leadership team, including Sarah Schinckel, Doug Sauder, Lane Arthur, Willy Pell and especially Chris Winkler who has already been an amazing partner in this new adventure. The photo below is the two of us in front of ISG today on a hot Des Moines morning.
-
Avni Patel Thompson reacted on thisAvni Patel Thompson reacted on thisTwo big things happened recently: we closed our next fund at Magnify Ventures, and I welcomed a daughter! We're thrilled to share this feature from Fortune’s Emma Hinchliffe on the launch of our new fund: bit.ly/4eGkACg I’m incredibly energized about what’s ahead. We’re at the start of a generational wave — companies building for the care economy, longevity, and women’s health are starting to get the attention they deserve. I couldn't be more excited to back the founders defining these categories — at a moment when AI is changing what's possible. Having a daughter a few weeks ago has sharpened everything. During 3am feedings, I'm thinking about what's coming — for working parents, for families — when visionary founders build from lived experience and the technology finally catches up. Endlessly thankful to Joanna Drake for her partnership, to the founders at Alix, Joy Parenting Club, Peak Health, Seven Starling, Quo Labs and beyond for letting us partner with you, and to our LPs for backing this vision. Our next fund is committed to the idea that care deserves to be built for, not just gotten through. That’s the wave we’re betting on — and the world I want my daughter to grow up in. Magnify Ventures, Joanna Drake, Fortune, Pivotal, The Holding Co., Marie Drez, Alexandra Mysoor, Hugh Tamassia, Mahima Chawla, Amy Cheslock, Sophia Richter, Tina Keshani, Wendy Johansson, Bismarck Lepe, Audrey Lo,Zhoujinyi, Daphne Li, Jeb Balise, Andrew Parker, Elisabeth Tarullo, Ben Hall
-
Avni Patel Thompson reacted on thisAvni Patel Thompson reacted on thisToday we're announcing Claude Corps, a $150M program that places 1,000 early-career fellows at 300+ nonprofits around the U.S. Our data indicates the early-career ladder is softening while thousands of under-resourced nonprofits are looking for support to safely use AI for their missions. Claude Corps is an initial effort to help address these needs. We are building this program and platform to be scaled and open-sourced, especially as the evidence comes in that it works for both fellows and host orgs. Grateful to CodePath and Social Finance for partnering with us on this ambitious effort. On a personal note, initiatives like Claude Corps are why I joined Anthropic. We are heading towards an important moment in the AI transition, so it's energizing to take actions in ways that align to our values as a public benefit corporation. Please encourage those in your network to apply: early-career talent (over 18 yrs old with less than 2 years of full-time work experience, STARs and recent grads) + nonprofits who are already using Claude! https://lnkd.in/gRU6XSjc
Experience & Education
Publications
-
Making Bad Analytics Good in China
Harvard Business Review Blog
Our 'on the ground' persective on how to turn bad data into usable analytics when working in China
Other authorsSee publication -
5 Things They Don't Tell You About Working in China
Harvard Business Review Blog
Languages
-
English
-
-
French
-
-
Hindi
-
-
Gujarati
-
-
Chinese (Mandarin)
-
Recommendations received
1 person has recommended Avni
Join now to viewOther similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content